A video may include one or more moments that are of interest to users. Analyzing content of a video to detect such moments may be computationally expensive. For example, decoding video frames within a video and analyzing the video frames may be computationally expensive.